1. Field of the Invention.
The present invention relates generally to the field of photonic devices. More specifically, the present invention discloses a monolithic optoelectronic TWE-component structure providing low insertion loss for high-speed optical signal transmission.
2. Statement of the Problem.
Low cost, small size, and high-performance indium phosphide (InP) based Mach-Zehnder modulators (MZMs) are widely known in the fiber optics photonic component industry. Two challenges exist that limit MZM performance: optical insertion loss and baud rate.
Optical insertion loss is defined as the power of the light coming out of the device divided by the power of the light injected into the device, usually expressed as the negative of the value in decibels (dB). The coupling of light onto and off of the chip usually involves bulk optical components and/or optical fiber waveguides, which have limitations on how small the light beam can be focused. On the chip, the light is guided by semiconductor epitaxial layers and photo-lithographically defined waveguides. For practical and cost reasons, the light beam on the chip has limitations on how large its diameter can be. This mismatch between the large off-chip beam and small on-chip beam gives rise to a coupling loss and is the major contributor to high loss in InP-based MZMs specifically and photonic devices in general. Coupling losses can be up to 4 dB or more for each of the input and output.
The baud rate of an MZM is limited by its radio frequency (RF) bandwidth. In an InP MZM, an electrical signal is applied to an electrode, which consequently switches the modulator between optical states. The bandwidth of the MZM is limited by the capacitance or impedance of the electrode, and the matching between the velocity of the optical wave and the electrical wave. A typical baud rate for an MZM with industrially acceptable performance is 10 GBd.
Both optical insertion loss problem and the baud rate problem have been solved separately. For insertion loss, a spot-size converter (SSC) has been developed which allows the size of the light beam on the semiconductor chip to expand to a size which is matched to off-chip bulk optics, improving the coupling loss to 0.8 dB. FIG. 1 shows a prior-art SSC 10 having a large mode 11 at the chip facet for coupling to bulk optics, and a small mode 12 at the end suitable for use on the chip. Note that the facet mode is present all the way to the top of the semiconductor. Note also that the prior art SSC 10 shown in FIG. 1 uses a selective etching and enhanced regrowth procedure to produce a core which varies in thickness in order to reduce the size of the mode as it propagates in the z direction. However in the prior art, the overclad above the core is of substantially uniform thickness. The SSC has been monolithically integrated with a MZM chip.
For the baud rate problem, a travelling wave electrode design has been developed which eliminates the capacitance and velocity matching limitation of the electrode. FIG. 2 is a plan view schematic of prior art travelling wave electrodes on a MZM 20. The black lines indicate, from left to right, an input waveguide 21, a splitting device 22, two MZ waveguides 23 and 24 in parallel, a combining device 25, and an output waveguide 26. The large grey areas indicate the travelling wave electrode 27 which periodically makes contact to the waveguides using t-shaped branches. Baud rates of 40 GBd have been demonstrated and a capability of extending the baud rate to 80 GBd has also been shown.
FIG. 3 is a cross-sectional view through the x-y plane in the MZ waveguide electrode region of the prior art SSC MZM 10 (left), and the prior art travelling wave MZM 20 (right). The layer with the horizontal lines indicates the semiconductor waveguide core 31. The dark grey layer on top indicates the metal of the electrode 32. In between the waveguide core 31 and the electrode 32 is a semiconductor layer 33. In the SSC MZM 10, this layer 33 must be thick to accommodate the large facet mode (as in FIG. 1), and lightly doped to minimize optical loss. In the travelling wave MZM 20, the layer 33 must be thin and highly doped to minimize RF loss.
Despite these separate advances in InP MZM technology, however, it has not been possible to integrate both improvements to simultaneously achieve both low insertion loss and high baud rate. There are two fundamental incompatibilities that are responsible. First, the travelling wave electrode requires that the layer of semiconductor material (referred to as the “overclad”) between the metal of the electrode on top and the guiding core underneath to be as thin as possible, in order to minimize RF loss. Contrarily, the SSC requires the overclad to be thick in order to allow the on-chip optical beam to expand sufficiently to match the bulk off-chip optics.
Second, the travelling wave electrode requires that the overclad be doped sufficiently to provide a highly conducting material to minimize the resistance. Contrarily, the SSC requires a low-doped or undoped overclad, since doped material (especially p-type doping) induces significant optical losses. These are hereafter referred to as the first and second incompatibilities, respectively.
3. Solution To The Problem.
The present invention addresses the two fundamental incompatibilities discussed above that have heretofore prevented the monolithic integration of an SSC and conventional travelling wave photonic devices, without sacrificing the benefits of each. In particular, the present invention employs a process of selective etching and enhanced regrowth (SEER) to create an overclad layer that is both: (1) thicker in the SSC region; and (2) highly doped in the travelling wave region of the device and undoped in the SSC region.